Intestinal Obstruction (Hirschsprung’s Disease & Intussusception) Brig Mushahid Aslam.
Clark, AD (2018) Mortality bene ts and intussusception...
-
Upload
duongtuong -
Category
Documents
-
view
215 -
download
0
Transcript of Clark, AD (2018) Mortality bene ts and intussusception...
Clark, AD (2018) Mortality benefits and intussusception risks of ro-tavirus vaccination in low- and middle-income countries. PhD (re-search paper style) thesis, London School of Hygiene & TropicalMedicine. DOI: https://doi.org/10.17037/PUBS.04651167
Downloaded from: http://researchonline.lshtm.ac.uk/4651167/
DOI: 10.17037/PUBS.04651167
Usage Guidelines
Please refer to usage guidelines at http://researchonline.lshtm.ac.uk/policies.html or alterna-tively contact [email protected].
Available under license: http://creativecommons.org/licenses/by-nc-nd/2.5/
http://researchonline.lshtm.ac.uk/4651167/http://dx.doi.org/10.17037/PUBS.04651167http://researchonline.lshtm.ac.uk/policies.htmlmailto:[email protected]
1
Mortality benefits and intussusception risks of rotavirus vaccination in low- and middle-income countries
ANDREW DAVID CLARK
Thesis submitted in accordance with the requirements for the degree of
Doctor of Philosophy of the University of London
December 2018
Department of Health Services Research and Policy
Faculty of Public Health and Policy
LONDON SCHOOL OF HYGIENE & TROPICAL MEDICINE
Funded in part by research grants from the World Health Organization and Bill and Melinda Gates Foundation
Supervisor: Professor Colin Sanderson
2
Signed Declaration
I Andrew David Clark, confirm that the work presented in this thesis is my own. Where
information has been derived from other sources, I confirm that this has been indicated in the
thesis.
Andrew David Clark
3
Abstract
Infant rotavirus vaccines have led to substantial reductions in rotavirus gastroenteritis (RVGE)
hospital admissions and costs, but some studies have reported an elevated risk of
intussusception, a rare bowel disorder, in vaccinated infants. The aim of this thesis is to
quantify the potential mortality benefits and intussusception risks of alternative rotavirus
vaccination schedules in 135 low- and middle-income countries (LMICs).
The thesis begins with an introduction to the topic and background to the literature and
concludes with some final reflections on the research and its relevance for informing national
decisions about vaccine safety and optimal scheduling of rotavirus vaccines. The main body
of the thesis includes a series of research papers which address specific topics relevant to the
estimation of mortality benefits and intussusception risks. These include methods for
estimating: RVGE deaths
4
Acknowledgements
I would like to thank Colin Sanderson for his encouragement and expert guidance over the
years. I also thank Mark Jit and Ulla Griffiths for providing helpful suggestions on earlier
drafts. I thank Emma, Ollie and Tom for their support and for inspiring me to get this work
finished.
5
Table of contents
Table of abbreviations...6
Chapter 1: Introduction.....8
Chapter 2: Review of the literature.............12
Chapter 3: Aim and objectives of thesis......31
Chapter 4: Estimation of rotavirus deaths in children aged
6
Table of abbreviations
AEFI Adverse Events Following Immunization AD Anderson-Darling statistic AIC Akaikes Information Criterion AIIMS All India Institute of Medical Science ATP According-To-Protocol BCG Bacillus Calmette-Gurin vaccine BCL Brighton Collaboration Level BIC Bayesian Information Criterion BMGF Bill and Melinda Gates Foundation BRV-PV Bovine Rotavirus Pentavalent Vaccine CDC Centers for Disease Control and Prevention CEA Cost-Effectiveness Analysis CFR Case Fatality Ratio CHERG Child Health Epidemiology Reference Group of UNICEF and the WHO CV Cramer-von Mises statistic DHS Demographic and Health Surveys DIC Deviance Information Criterion DTP1 Diphtheria-Tetanus-Pertussis vaccine dose 1 DTP2 Diphtheria-Tetanus-Pertussis vaccine dose 2 DTP3 Diphtheria-Tetanus-Pertussis vaccine dose 3 EIA Enzyme Immunoassay FEC Finnish Extension Trial FUP Follow-up FVI Fully Vaccinated Infants GAVCS Global Advisory Committee on Vaccine Safety GAVI Global Alliance for Vaccines and Immunization GBD Global Burden of Disease Project GE Gastroenteritis GEMS Global Enteric Multicenter Study
GMC Geometric Mean Concentration GNI Gross National Income GRSN WHO-coordinated Global Sentinel Site Rotavirus Surveillance Network
HIC High-Income Country ICD International Classification of Diseases IGME UN Inter-agency Group for Child Mortality Estimation
IHME Institute of Health Metrics and Evaluation IV Intravenous rehydration iVE Instantaneous Vaccine Efficacy IVIR-AC Immunization and Vaccines-related Implementation Research Advisory Committee KDE Kernel Density Estimation KS Kolmogorov-Smirnov statistic LLR Lanzhou Lamb Rotavirus vaccine LMIC Low- and Middle-Income Country LSHTM London School of Hygiene and Tropical Medicine MAE Mean Absolute Error
7
MAL-ED Malnutrition and Enteric Disease Study MCEE Maternal Child Epidemiology Estimation Group MCMC Markov Chain Monte Carlo MCRI Murdoch Childrens Research Institute MCV1 Measles-Containing Vaccine dose 1 Meas1 Measles-Containing Vaccine dose 1 MICS Multiple Indicator Cluster Surveys MLE Maximum Likelihood Estimation MMR Measles Mumps and Rubella vaccine MSD Moderate-to-Severe Diarrhoea NIH National Institutes of Health NLS Non-Linear Least Squares NRSN Indian National Hospital Rotavirus Surveillance Network OPV Oral Polio Vaccine ORS Oral Rehydration Salts/Solution PCR Polymerase Chain Reaction PRISMA Preferred Reporting Items for Systematic Reviews and Meta-Analyses RCH Royal Childrens Hospital, Melbourne RCT Randomised Controlled Trial REST Rotavirus Efficacy and Safety Trial RMSE Root Mean Squared Error RR Relative Risk RV3-BB Rotavirus Vaccine based on G3 P6 strain (Bishop and Barnes) RVGE Rotavirus Gastroenteritis SAE Severe Adverse Events SAGE Strategic Advisory Group of Experts SCCS Self-Controlled Case Series U5 Under-five years of age UNICEF United Nations International Children's (Emergency) Fund UNPOP United Nations Population Division USAID United States of America International Development VAPP Vaccine Associated Paralytic Poliomyelitis VIMC Vaccine Impact Modelling Consortium WAIFW Who Acquires Infection From Whom WHO World Health Organization WUENIC WHO and UNICEF Estimates of National Immunization Coverage
8
1.0 Chapter 1 Introduction
1.1 Benefits and risks of live oral rotavirus vaccines
Rotavirus is an important cause of gastroenteritis (GE) in children aged
9
two doses of Rotarix with DTP1 and DTP2, or three doses of RotaTeq with DTP1,
DTP2, and DTP3 as per current WHO recommendations. The actual age of
administration varies between countries due to differences in national schedules
(target ages for DTP) and differences in the timeliness of vaccination (9). The standard
infant schedules recommended by WHO have demonstrated high and durable efficacy
in low mortality countries but modest efficacy in higher mortality settings (10). This
has stimulated interest in the potential value of a booster dose given with the first dose
of measles-containing vaccine (MCV1, referred to hereafter as Meas1)(11) or a birth
dose given at the same time as Bacillus Calmette-Gurin (BCG)(12). A birth dose has
the potential to prevent disease that occurs very early in life, while a booster dose has
the potential to mitigate the effects of waning rotavirus vaccine protection, a
phenomenon observed in several high mortality settings (10). Birth doses also have
the potential to reduce the number of excess (vaccine-related) intussusception cases
by administering the first dose earlier in life, when the background risk of
intussusception is lower. The optimal number and timing of doses (concurrent with
different combinations of BCG, DTP1, DTP2, DTP3 and Meas1) will depend on
several criteria, including the balance of benefits to risks i.e. number of RVGE deaths
averted per excess intussusception death.
1.3 Scope of the thesis
The aim of this thesis is to quantify the potential mortality benefits (averted RVGE
deaths
10
years (Chapter 5); vaccine coverage and timeliness (Chapter 6); rotavirus vaccine
efficacy and waning (Chapter 7); and, intussusception incidence, age distributions and
case fatality ratios - CFRs (Chapter 8). The final research paper (Chapter 9) brings
together this evidence and uses a national vaccine decision support model to estimate
the potential rotavirus mortality benefits and intussusception risks of 18 possible
rotavirus vaccination schedules in 135 LMICs. The thesis concludes with some final
reflections on the research and its relevance for informing national decisions about
vaccine safety and optimal scheduling of rotavirus vaccines (Chapter 10).
This is a research-paper thesis rather than the conventional book style. Six of the
chapters are full research papers. Two have been published in peer-reviewed journals
(Chapters 4 & 6) and the remaining four (Chapters 5, 7, 8 and 9) have been prepared
for submission. For consistency between published and unpublished research papers,
references are listed at the end of each chapter, and the numbering of tables and figures
is restarted at the beginning of each chapter. Each research-paper chapter begins with
a short section describing how the paper contributes to the overall aim and objectives
of the thesis. I have also described my independent academic contribution to each
paper. This is important to clarify because all papers include contributions from
others. Relevant funding and ethical approvals are also described.
Appendices to Chapters 5, 8 and 9 are included at the end of each chapter for ease of
referencing. Other appendices are included at the end of the thesis, either because they
contain optional background material (Appendices 1-5, 10), or because they describe
analyses that were primarily done by others (Appendices 6-9).
11
List of references (Chapter 1)
1. Clark A, Black R, Tate J, Roose A, Kotloff K, Lam D, et al. Estimating global, regional and national rotavirus deaths in children aged
12
2.0 Chapter 2 Review of the literature
2.1 Rotavirus gastroenteritis
Gastroenteritis (GE), characterised by diarrhoea, abdominal cramps, nausea and
sometimes vomiting, is usually diagnosed when a child experiences three or more
loose stools, or vomiting, within a 24-hour period (1). Without adequate fluid
replenishment GE can quickly lead to dehydration, electrolyte imbalance, metabolic
acidosis, shock and death (2). Descriptions of GE exist as far back as the earliest
records of human civilisation (3) and its role in child mortality is well documented -
described in 1940s London as one of the most fatal diseases of infancy in our
capital(4).
In 1973 Ruth Bishop, Geoffrey Davidson, Ian Holmes and Brian Ruck identified a
high volume of particles of a new virus in the faeces of children admitted to the Royal
Childrens Hospital (RCH) in Melbourne, Australia. The wheel-like structure seen
under an electron microscope was the inspiration for the name rotavirus (rota is latin
for wheel) (5). Prior to this discovery, the causative agents of GE were poorly
understood. Rotavirus has since been detected in numerous studies around the world
and is now recognised as a leading global cause of GE hospitalisations in children
aged
13
Following a brief period of protection from maternal antibodies, almost every child in
the world, irrespective of where they live, will be infected with rotavirus at least once
before their fifth birthday. The mode of transmission is thought to be faecal-oral, and
the incubation period (interval between exposure and onset of symptoms) is usually
less than 48 hours (2). Natural or wild type rotavirus infections (asymptomatic or
symptomatic) have been shown to provide some protection against subsequent
moderate-to-severe disease. Important birth cohort studies have been conducted in
Mexico and India. In Mexico, two prior infections provided complete protection
against subsequent moderate-to-severe disease (10). In India however, two infections
provided only 57% protection (and only 79% after three) (11). In both settings, natural
infections provided limited protection against subsequent asymptomatic infections
and mild disease. Thus frequent reinfections are probably very common and provide
the basis for continued circulation. A study in England detected rotavirus in the stools
of healthy individuals in all age groups (12).
Two point scoring systems have been used to determine the severity of RVGE in
randomised controlled trials (RCTs), the 20-point Vesikari system and the 24-point
Clark system (13). Points are awarded for the presence and severity of different
symptoms e.g. duration of vomiting, duration of diarrhoea, rectal temperature, signs
of dehydration etc. Nearly all RCTs of rotavirus vaccines define severe RVGE as 11-
20 points on the Vesikari scale. Some older trials use a Clark score of 16-24 points.
These two scores have been shown to correlate poorly with one another when
estimating the proportion of RVGE episodes defined as severe (13, 14).
In the scientific literature, the most commonly reported rotavirus disease burden
indicator is the rotavirus-positive proportion among GE hospital admissions aged
14
vomiting and diarrhoea, intravenous rehydration (IV) is required in more serious cases
(16).
Estimates of the number of RVGE deaths
15
A good deal of pre- and post-licensure evidence has now been accumulated on the
efficacy and safety of Rotarix and RotaTeq. Both vaccines have demonstrated
>90% efficacy against severe RVGE episodes in low mortality settings. However, the
combined estimate of efficacy from higher mortality countries (Bangladesh, Vietnam,
Ghana, Kenya and Mali) is only around 67% in the first year of life and 34% in the
second year of life (24). The reason for lower and less durable efficacy in these
settings is unclear but has been linked to factors including poorer nutritional status
and more frequent exposure to a wider range of enteric pathogens (25). Both vaccines
have demonstrated clinical cross-protection to the major strains not included in the
vaccines (26), and thus similar efficacy in different settings. It is difficult to compare
the two vaccines directly as different case definitions and severity scales were used in
some trials (13) and the two vaccines have never been directly compared in a head-
to-head RCT within the same study population. However, the post-licensure
experience of countries that have used both Rotarix and RotaTeq does not suggest
any material difference in impact (27).
In recent years, several manufacturers have emerged from LMICs. In China, LLR
(Lanzhou lamb rotavirus vaccine - Lanzhou Institute of Biological Products) is based
on a single G10P[12] lamb rotavirus strain, and has been sold in the private market
since 2000. A case control study of LLR in China reported effectiveness of 43% for
a single dose administered between 2 and 35 months of age (28). In Vietnam,
Rotavin (POLYVAC-Vietnam) is based on an attenuated human G1P[8] strain
isolated from a Vietnamese child. This vaccine was licensed for use after
demonstrating comparable immunogenicity to Rotarix among Vietnamese infants
(29). In India, two vaccines have been developed. ROTAVAC (Bharat Biotech
International, India) was recently WHO pre-qualified for global use (30). This is based
on a naturally attenuated human-bovine (cow) strain (II6E) isolated from an Indian
infant by Maharaj K Bahn and colleagues at the All India Institute of Medical Sciences
(AIIMS) in Delhi (31). This vaccine demonstrated 54% efficacy against severe RVGE
in Indian infants and is priced at less than $3 per three-dose course. This is
considerably lower than the price of Rotarix and RotaTeq. A bovine-human
reassortant pentavalent rotavirus vaccine (BRV-PV) called ROTASIIL (Serum
Institute of India) is also licensed for use in India. This has demonstrated 38% efficacy
against severe RVGE in India, and 65% efficacy in Niger (32, 33). This vaccine may
also be WHO pre-qualified for global use soon.
16
Several others rotavirus vaccines are also in the pipeline, including neonatal and non-
replicating injectable vaccines (34, 35). An oral neonatal vaccine, RV3-BB (Murdoch
Childrens Research Institute - MCRI, Australia) has shown to be efficacious in
Indonesia when administered as a three-dose series (36). This has the potential to
increase vaccination coverage and potentially reduce vaccine-related intussusception
events by allowing the first dose to be administered earlier in life when the
background rate of intussusception is low. Other vaccines in the pipeline include non-
replicating injectable vaccines (35). These could prove to be safer and more
efficacious than existing live oral vaccines, but more clinical evidence will be needed
to confirm this.
2.3 Rotavirus vaccination and intussusception
Intussusception is the main cause of bowel obstruction in children aged
17
associated with an increased risk of intussusception in some settings (38, 39). While
the scale of risk associated with Rotarix and RotaTeq is believed to be smaller
than the risk observed for RotaShield, a true comparison is not possible because
both vaccines have been administered within age windows designed to avoid the
background peak age of intussusception. Many of the vaccines administered in the
RotaShield programme were administered as part of a catch-up campaign, so were
administered to older infants when the background incidence of intussusception was
high (40). To limit the scale of potential vaccine-related intussusception cases the two
major vaccine manufacturers have recommended different age restrictions tailored to
their own vaccines. The aim of these age restrictions is to ensure the vaccine is
administered earlier in life, when the background incidence of intussusception is
lower. The WHO harmonised the different manufacturers age restrictions and
recommended administration of the first dose before 15 weeks of age and the final
dose before 32 weeks of age (24). Following a modelling analysis in 2012, the WHO
revised their recommendation to allow countries to remove age restrictions in
countries where the benefits of later vaccination would greatly exceed the risks (24,
41).
An excess or vaccine-related case of intussusception is defined by WHO as an adverse
event following immunization (AEFI). Intussusception is a serious adverse event
because it has the potential to lead to hospitalisation and death. Serious adverse events
are included within the wider spectrum of all severe adverse events (SAE). SAEs also
include severe reactions that are not life-threatening. The distinction between
serious and severe is important; serious is a regulatory term whereas severe is not
(42).
Different study designs have been used to detect any potential relative risk of
intussusception following vaccination, but because intussusception is such a rare
event, these studies are often not powered for reliable detection of an increase in risk.
The self-controlled case series (SCCS) methodology is considered to be relatively
reliable and has been widely used (5). In this method, children with intussusception
act as their own controls. The risk of intussusception is calculated for the period of
hypothesised elevated risk (i.e. 21 days following vaccination) and then compared to
the risk of intussusception in all other periods, with appropriate adjustment for age
(43). Intussusception risk is expressed as the relative incidence (RI) or relative risk
(RR) compared to the expected background incidence in the absence of vaccination.
A meta-analysis of Rotarix studies by Stowe et al found pooled RI estimates of 2.4
18
(95% confidence interval 1.5 - 3.8) and 1.8 (1.3 - 2.4) in the 21 day period after the
first and second doses, respectively (44). Similar risks have also been reported for
RotaTeq in the USA (39) and Australia (38). The estimates for Rotarix and
RotaTeq in Australia were equivalent to one additional case in every 14,000-20,000
vaccinated infants, but studies from other settings have reported a lower level of risk,
and none have reported a risk as high as RotaShield (one in every 5,000-10,000
vaccinated infants). Encouragingly, a recent SCCS study of Rotarix in Africa found
no elevated risk of intussusception in the first 1-7 days after dose 1 (RI 0.30, 95% CI
0.0 - 1.0) or dose 2 (RI 0.8, 95% CI 0.2 1.7) and no elevated risk 8-21 days after
dose 1 (RI 1.0, 95% CI 0.3 2.3) or dose 2 (RI 0.7, 95% CI 0.4 1.2). This was a
multi-site study including infants from seven countries (Ethiopia, Ghana, Kenya,
Malawi, Tanzania, Zambia, and Zimbabwe) (45).
2.4 Published studies evaluating the benefits and risks of rotavirus vaccination
In 2012, Patel et al estimated the potential mortality benefits and intussusception risks
of introducing live oral rotavirus vaccines into the 2010 birth cohort of 158 countries
(24). This study updated an earlier analysis conducted in 2009 based on 117 countries
(46). The 2012 study found that with full adherence to the manufacturers age
restrictions, universal introduction of rotavirus vaccination would prevent 156,000
RVGE deaths and cause 253 deaths (benefit-risk ratio of ~600:1). Without age
restrictions, rotavirus vaccines were estimates to prevent 203,000deaths and cause 547
deaths (benefit-risk ratio of ~370:1). The study therefore found that removing age
restrictions from a standard infant schedule co-administered with DTP would prevent
an additional ~47,000 RVGE deaths and potentially cause an additional ~300
intussusception deaths each year (incremental benefit-risk ratio of 154:1)(24). This
study informed a WHO recommendation to remove the manufacturers age
restrictions for vaccination given that the benefits of preventing additional rotavirus
mortality from later vaccination greatly exceeded the intussusception risks (41). The
2012 publication (Appendix 1) and WHO position paper (Appendix 2) are available
in the list of appendices.
Several other benefit-risk analyses have also been published. A multi-country analysis
for all countries in Latin America estimated a benefit-risk ratio of 395:1 (47), while
in a study in Brazil and Mexico the estimate was 260:1 (48). The benefit-risk ratio
was estimated to be 88:1 in England (18), 273:1 in France (49), 77:1 in the USA (50)
and 366:1 in Japan (51). Other high income countries have calculated benefits and
19
risks for hospitalisations, but not mortality (38). In higher income countries, mortality
from both rotavirus and intussusception is very rare, and other criteria become more
important. In England, Clark et al estimated that Rotarix would cause one additional
intussusception admission in every 18,551 vaccinated English infants (5th and 95th
percentiles, 6,728 - 93,952), equivalent to 35 additional intussusception admissions
each year. In contrast, it was estimated that each year the vaccine prevented three
rotavirus deaths, 13,000 rotavirus admissions, 27,000 rotavirus emergency visits and
74,000 rotavirus GP consultations in children aged
20
influence on vaccine impact estimates. In contrast, waning duration of protection was
shown to have an important influence in some of the analyses. Only one (2%) of the
models explicitly modelled the natural history of disease and associated herd effects
(54). Postma et al compared estimates from three models and attributed differences in
results to assumptions about dose-specific vaccine efficacy, waning duration of
protection and the level of immunity acquired from natural infections (55). The
timeliness of vaccination was shown to be influential in a cost-effectiveness study by
Clark et al in Peru. This analysis showed that ignoring delays and assuming on-time
vaccination would over-estimate health benefits by 4% (56). An analysis of delays in
45 other countries by Clark and Sanderson showed that Perus immunization
programme is relatively timely compared to others, so this error is likely to be greater
in countries with more pronounced delays (57).
Unlike static cohort models, transmission dynamic models are able to predict the
number of susceptible, infectious and immune individuals over time. These models
are also able to capture the interplay between immunity acquired from vaccination
and immunity acquired from repeated natural (wild type) infections. Models of this
kind typically require assumptions to be made about the number of individuals
exposed by each infectious individual (also known as the basic reproductive number
or R0), the duration of immunity acquired from natural infections, and further
assumptions about who acquires infection from whom (WAIFW). Pitzer et al
described five rotavirus transmission dynamic models that were each calibrated to the
same age-specific RVGE incidence data from England and Wales (58). All five
models simulated the flow of groups of individuals into different compartments
(health states) over time using differential equations. A pivotal study by Velasquez et
al was used to inform estimates of protection from 1 up to 4 natural infections against
subsequent infections and disease episodes (10). Estimates of R0 varied considerably
between the five models, ranging from ~1 to 26 secondary exposures per infectious
individual. Discrepancies between the model predictions reflected uncertainties in the
age-specific risk of RVGE infections, and the duration of natural and vaccine-induced
immunity. However, over the long-term (5 years post-vaccination), all of the models
predicted impact among children aged
21
better fit to the local data. The authors found a marginal role for herd effects in
explaining overall impact (~1% of the total long-term impact in children aged
22
Graphical depiction of the trolley problem
Source: https://i0.wp.com/moralarc.org/wp-content/uploads/2015/04/trolley-problem.jpg?w=620
Another variation is that a person is pushed from a bridge into the path of the tram,
again saving five lives at the expense of one. The moral dilemma is a choice between
inaction and intervention. Utilitarianism (the greatest good for the greatest number)
would favour intervention (69). In the context of rotavirus vaccination this would
mean favouring schedule options that maximise the net number of deaths prevented,
irrespective of whether large number of intussusception deaths are caused in the
process. This raises important ethical considerations and is contrary to the public
health principle first do no harm (67). It also fails to consider uncertain
consequences that could be associated with taking action. For example, an increase in
high profile legal challenges and anti-vaccine sentiment could have an adverse effect
on the coverage of rotavirus vaccines, and potentially other vaccines. In England,
concerns about the safety of whole-cell pertussis and MMR (measles mumps and
rubella) vaccines have previously led to substantial short-term declines in coverage
(65). There is also some evidence that a death caused by action/intervention may be
perceived by individuals as worse than a death caused by inaction (70, 71).
Herbert Simon made a distinction between substantive rationality, choosing the
outcome with the maximum mathematical utility, and procedural rationality, allowing
decision makers to reject options that do not meet minimum standards (72). In terms
https://i0.wp.com/moralarc.org/wp-content/uploads/2015/04/trolley-problem.jpg?w=620
23
of rotavirus vaccination, this would imply a maximum level of acceptable risk, above
which the vaccination programme would not be socially acceptable. The risk
associated with RotaShield in the USA (more than one excess intussusception cases
in every 10,000 vaccinated infants) provides an important psychological benchmark
for what might be considered a maximum level of acceptable risk. Fine and Clarkson
have argued that the level of acceptable risk will differ depending on whether the
choice is made by individuals (more likely to choose lower uptake) or public health
decision-makers representing the community as a whole (more likely to choose higher
uptake) (73). Another benchmark that could be used to inform a socially acceptable
risk for rotavirus vaccines is the level of risk that has been accepted for other vaccines.
However, combining this evidence is not straightforward. The WHO provides
reported reaction rates for each vaccine formulation but the spectrum of possible
adverse effects is broad and the uncertainty intervals around the risks are wide. For
BCG vaccine, the risk of disseminated BCG disease (fatal in 50% of cases) is reported
to be less than one in every 200,000 vaccinated infants. For the first dose of oral polio
vaccine (OPV) the risk of vaccine associated paralytic poliomyelitis (VAPP) is one in
every 750,000 vaccinated infants. For Measles and DTP vaccines, rates of febrile
seizures are relatively common (one in every ~3000 doses) but these are rarely fatal
(42, 74). Resnik has argued that the maximum level of acceptable risk should not
exceed the maximum risk of death for high risk forms of paid labour, such the
mortality risks among fishermen, loggers and extraction workers. He went on to
suggest this as one possible approach for determining the maximum acceptable risk
among paid volunteers in clinical trials. The maximum acceptable risk of a serious
adverse event could then be derived by combining the maximum acceptable mortality
risk with the CFR for the serious adverse event in question (75). This approach has
obvious limitations if applied to the example of rotavirus vaccination because the
focus is on adults and paid participation.
For rotavirus vaccination, the maximum acceptable risk will be inextricably linked to
the scale of potential benefits and for this reason, it would be very difficult for national
decision-making committees to set universal thresholds for maximum acceptable risk.
A minimum threshold for the balance of benefits to risks (minimum benefit-risk ratio)
could however be developed, and may lead to more consistent decision-making across
vaccines. The Global Advisory Committee on Vaccines Safety (GACVS) and
Strategic Advisory Group of Experts (SAGE) are the principal advisory groups to
WHO on issues around the safety and acceptability of rotavirus vaccines, and
ultimately the committee members will have to make value judgements and
24
recommendations informed by the best available evidence on benefits and risks, as
well as other criteria including costs, cost-effectiveness and operational feasibility
(76).
25
List of references (Chapter 2)
1. Majowicz SE, Hall G, Scallan E, Adak GK, Gauci C, Jones TF, et al. A common, symptom-based case definition for gastroenteritis. Epidemiol Infect. 2008;136(7):886-94.
2. Groome MJ, Zell ER, Solomon F, Nzenze S, Parashar UD, Izu A, et al. Temporal
Association of Rotavirus Vaccine Introduction and Reduction in All-Cause Childhood Diarrheal Hospitalizations in South Africa. Clin Infect Dis. 2016;62 Suppl 2:S188-95.
3. Lim ML, Wallace MR. Infectious diarrhea in history. Infect Dis Clin North Am.
2004;18(2):261-74. 4. Gairdner P. Infantile diarrhoea: An Analysis of 216 Cases with Special Reference to
Institutional Outbreaks. Arch Dis Child. 1945;20(101):22-7. 5. Bishop R. Discovery of rotavirus: Implications for child health. J Gastroenterol
Hepatol. 2009;24 Suppl 3:S81-5. 6. Clark A, Black R, Tate J, Roose A, Kotloff K, Lam D, et al. Estimating global,
regional and national rotavirus deaths in children aged
26
14. Aslan A, Kurugol Z, Cetin H, Karakaslilar S, Koturoglu G. Comparison of Vesikari and Clark scales regarding the definition of severe rotavirus gastroenteritis in children. Infect Dis (Lond). 2015;47(5):332-7.
15. Kotloff KL, Blackwelder WC, Nasrin D, Nataro JP, Farag TH, van Eijk A, et al. The
Global Enteric Multicenter Study (GEMS) of diarrheal disease in infants and young children in developing countries: epidemiologic and clinical methods of the case/control study. Clin Infect Dis. 2012;55 Suppl 4:S232-45.
16. WHO. World Health Organization. Diarrhoeal Disease Fact Sheet. Available at:
http://www.who.int/en/news-room/fact-sheets/detail/diarrhoeal-disease [accessed 12th August 2018]. 2018.
17. Bilcke J, Van Damme P, Van Ranst M, Hens N, Aerts M, Beutels P. Estimating the
incidence of symptomatic rotavirus infections: a systematic review and meta-analysis. PLoS One. 2009;4(6):e6060.
18. Clark A. JM, Andrews N., Atchison C., Edmunds J., Sanderson C., . Evaluating the
potential risks and benefits of infant rotavirus vaccination in England. Vaccine. 2014.
19. Schwartz JL. The first rotavirus vaccine and the politics of acceptable risk. Milbank
Q. 2012;90(2):278-310. 20. Ruiz-Palacios GM, Perez-Schael I, Velazquez FR, Abate H, Breuer T, Clemens SC,
et al. Safety and efficacy of an attenuated vaccine against severe rotavirus gastroenteritis. N Engl J Med. 2006;354(1):11-22.
21. Vesikari T, Matson DO, Dennehy P, Van Damme P, Santosham M, Rodriguez Z, et
al. Safety and efficacy of a pentavalent human-bovine (WC3) reassortant rotavirus vaccine. N Engl J Med. 2006;354(1):23-33.
22. Ward RL, Bernstein DI. Rotarix: a rotavirus vaccine for the world. Clin Infect Dis.
2009;48(2):222-8. 23. Clark HF, Offit PA, Plotkin SA, Heaton PM. The new pentavalent rotavirus vaccine
composed of bovine (strain WC3) -human rotavirus reassortants. Pediatr Infect Dis J. 2006;25(7):577-83.
24. Patel MM, Clark AD, Sanderson CF, Tate J, Parashar UD. Removing the age
restrictions for rotavirus vaccination: a benefit-risk modeling analysis. PLoS Med. 2012;9(10):e1001330.
25. Bhandari N, Rongsen-Chandola T, Bavdekar A, John J, Antony K, Taneja S, et al.
Efficacy of a monovalent human-bovine (116E) rotavirus vaccine in Indian infants: a randomised, double-blind, placebo-controlled trial. Lancet. 2014.
26. Vesikari T. Rotavirus vaccination: a concise review. Clin Microbiol Infect. 2012;18
Suppl 5:57-63. 27. Immergluck LC, Parker TC, Jain S, Laghaie E, Spandorfer P, Jerris RC, et al.
Sustained Effectiveness of Monovalent and Pentavalent Rotavirus Vaccines in Children. J Pediatr. 2016;172:116-20 e1.
http://www.who.int/en/news-room/fact-sheets/detail/diarrhoeal-disease
27
28. Fu C, He Q, Xu J, Xie H, Ding P, Hu W, et al. Effectiveness of the Lanzhou lamb rotavirus vaccine against gastroenteritis among children. Vaccine. 2012;31(1):154-8.
29. Dang DA, Nguyen VT, Vu DT, Nguyen TH, Nguyen DM, Yuhuan W, et al. A dose-
escalation safety and immunogenicity study of a new live attenuated human rotavirus vaccine (Rotavin-M1) in Vietnamese children. Vaccine. 2012;30 Suppl 1:A114-21.
30. WHO. WHO. WHO prequalified vaccines. Last updated 8/8/2018. Available at:
https://extranet.who.int/gavi/PQ_Web/ [accessed 8th August 2018]. 2018. 31. Bhan MK, Glass RI, Ella KM, Bhandari N, Boslego J, Greenberg HB, et al. Team
science and the creation of a novel rotavirus vaccine in India: a new framework for vaccine development. Lancet. 2014.
32. Isanaka S, Guindo O, Langendorf C, Matar Seck A, Plikaytis BD, Sayinzoga-
Makombe N, et al. Efficacy of a Low-Cost, Heat-Stable Oral Rotavirus Vaccine in Niger. N Engl J Med. 2017;376(12):1121-30.
33. Kulkarni PS, Desai S, Tewari T, Kawade A, Goyal N, Garg BS, et al. A randomized
Phase III clinical trial to assess the efficacy of a bovine-human reassortant pentavalent rotavirus vaccine in Indian infants. Vaccine. 2017;35(45):6228-37.
34. Yen C, Tate JE, Hyde TB, Cortese MM, Lopman BA, Jiang B, et al. Rotavirus
vaccines: Current status and future considerations. Hum Vaccin Immunother. 2014;10(6).
35. Kirkwood CD, Ma LF, Carey ME, Steele AD. The rotavirus vaccine development
pipeline. Vaccine. 2017. 36. Bines JE, At Thobari J, Satria CD, Handley A, Watts E, Cowley D, et al. Human
Neonatal Rotavirus Vaccine (RV3-BB) to Target Rotavirus from Birth. N Engl J Med. 2018;378(8):719-30.
37. Mpabalwani EM, Mwenda JM, Tate JE, Parashar UD. Review of Naturally
Occurring Intussusception in Young Children in the WHO African Region prior to the Era of Rotavirus Vaccine Utilization in the Expanded Programme of Immunization. J Trop Pediatr. 2017;63(3):221-8.
38. Carlin JB, Macartney KK, Lee KJ, Quinn HE, Buttery J, Lopert R, et al.
Intussusception risk and disease prevention associated with rotavirus vaccines in Australia's National Immunization Program. Clin Infect Dis. 2013;57(10):1427-34.
39. Yih WK, Lieu TA, Kulldorff M, Martin D, McMahill-Walraven CN, Platt R, et al.
Intussusception risk after rotavirus vaccination in U.S. infants. N Engl J Med. 2014;370(6):503-12.
40. Simonsen L, Viboud C, Elixhauser A, Taylor RJ, Kapikian AZ. More on RotaShield
and intussusception: the role of age at the time of vaccination. J Infect Dis. 2005;192 Suppl 1:S36-43.
41. WHO. WHO position paper January 2013. Rotavirus. Weekly Epidemiological
Review (WER) Available at: http://wwwwhoint/wer/2013/wer8805/en/ [Accessed 7 Feb 2018]. 2013;88(5):4964.
https://extranet.who.int/gavi/PQ_Web/http://wwwwhoint/wer/2013/wer8805/en/
28
42. WHO. Module 3. Adverse events following immunization. Available at: http://www.who.int/vaccine_safety/initiative/tech_support/Part-3.pdf [accessed 13th August 2018].
43. Petersen I, Douglas I, Whitaker H. Self controlled case series methods: an
alternative to standard epidemiological study designs. BMJ. 2016;354:i4515. 44. Stowe J, Andrews N, Ladhani S, Miller E. The risk of intussusception following
monovalent rotavirus vaccination in England: A self-controlled case-series evaluation. Vaccine. 2016;34(32):3684-9.
45. Tate JE, Mwenda JM, Armah G, Jani B, Omore R, Ademe A, et al. Evaluation of
Intussusception after Monovalent Rotavirus Vaccination in Africa. N Engl J Med. 2018;378(16):1521-8.
46. Patel MM, Clark AD, Glass RI, Greenberg H, Tate J, Santosham M, et al.
Broadening the age restriction for initiating rotavirus vaccination in regions with high rotavirus mortality: benefits of mortality reduction versus risk of fatal intussusception. Vaccine. 2009;27(22):2916-22.
47. Desai R, Parashar UD, Lopman B, de Oliveira LH, Clark AD, Sanderson CF, et al.
Potential intussusception risk versus health benefits from rotavirus vaccination in Latin America. Clin Infect Dis. 2012;54(10):1397-405.
48. Patel MM, Lopez-Collada VR, Bulhoes MM, De Oliveira LH, Bautista Marquez A,
Flannery B, et al. Intussusception risk and health benefits of rotavirus vaccination in Mexico and Brazil. N Engl J Med. 2011;364(24):2283-92.
49. Lamrani A, Tubert-Bitter P, Hill C, Escolano S. A benefit-risk analysis of rotavirus
vaccination, France, 2015. Euro Surveill. 2017;22(50). 50. Desai R, Cortese MM, Meltzer MI, Shankar M, Tate JE, Yen C, et al. Potential
intussusception risk versus benefits of rotavirus vaccination in the United States. Pediatr Infect Dis J. 2013;32(1):1-7.
51. Ledent E, Lieftucht A, Buyse H, Sugiyama K, McKenna M, Holl K. Post-Marketing
Benefit-Risk Assessment of Rotavirus Vaccination in Japan: A Simulation and Modelling Analysis. Drug Saf. 2016;39(3):219-30.
52. Bilcke J, Beutels P. Reviewing the cost effectiveness of rotavirus vaccination: the
importance of uncertainty in the choice of data sources. Pharmacoeconomics. 2009;27(4):281-97.
53. Postma MJ, Jit M, Rozenbaum MH, Standaert B, Tu HA, Hutubessy RC.
Comparative review of three cost-effectiveness models for rotavirus vaccines in national immunization programs; a generic approach applied to various regions in the world. BMC Med. 2011;9:84.
54. Aballea S, Millier A, Quilici S, Caroll S, Petrou S, Toumi M. A critical literature
review of health economic evaluations of rotavirus vaccination. Hum Vaccin Immunother. 2013;9(6):1272-88.
55. Tu HA, Woerdenbag HJ, Kane S, Rozenbaum MH, Li SC, Postma MJ. Economic
evaluations of rotavirus immunization for developing countries: a review of the literature. Expert Rev Vaccines. 2011;10(7):1037-51.
http://www.who.int/vaccine_safety/initiative/tech_support/Part-3.pdf
29
56. Clark AD, Walker DG, Mosqueira NR, Penny ME, Lanata CF, Fox-Rushby J, et al.
Cost-effectiveness of rotavirus vaccination in peru. J Infect Dis. 2009;200 Suppl 1:S114-24.
57. Clark A, Sanderson C. Timing of children's vaccinations in 45 low-income and
middle-income countries: an analysis of survey data. Lancet. 2009;373(9674):1543-9.
58. Pitzer VE, Atkins KE, de Blasio BF, Van Effelterre T, Atchison CJ, Harris JP, et al.
Direct and indirect effects of rotavirus vaccination: comparing predictions from transmission dynamic models. PLoS One. 2012;7(8):e42320.
59. Park J, Goldstein J, Haran M, Ferrari M. An ensemble approach to predicting the
impact of vaccination on rotavirus disease in Niger. Vaccine. 2017;35(43):5835-41. 60. Rose J, Homa L, Meropol SB, Debanne SM, Bielefeld R, Hoyen C, et al. Health
impact and cost-effectiveness of a domestically-produced rotavirus vaccine in India: A model based analysis. PLoS One. 2017;12(11):e0187446.
61. Bennett A, Bar-Zeev N, Cunliffe NA. Measuring indirect effects of rotavirus
vaccine in low income countries. Vaccine. 2016;34(37):4351-3. 62. Yen C, Armero Guardado JA, Alberto P, Rodriguez Araujo DS, Mena C, Cuellar E,
et al. Decline in rotavirus hospitalizations and health care visits for childhood diarrhea following rotavirus vaccination in El Salvador. Pediatr Infect Dis J. 2011;30(1 Suppl):S6-S10.
63. Burnett E, Jonesteller CL, Tate JE, Yen C, Parashar UD. Global Impact of Rotavirus
Vaccination on Childhood Hospitalizations and Mortality From Diarrhea. J Infect Dis. 2017;215(11):1666-72.
64. Ngabo F, Tate JE, Gatera M, Rugambwa C, Donnen P, Lepage P, et al. Effect of
pentavalent rotavirus vaccine introduction on hospital admissions for diarrhoea and rotavirus in children in Rwanda: a time-series analysis. Lancet Glob Health. 2016;4(2):e129-36.
65. Bennett A, Pollock L, Jere KC, Pitzer VE, Parashar U, Tate JE, et al. Direct and
possible indirect effects of vaccination on rotavirus hospitalisations among children in Malawi four years after programmatic introduction. Vaccine. 2018.
66. Mpabalwani EM, Simwaka CJ, Mwenda JM, Mubanga CP, Monze M, Matapo B, et
al. Impact of Rotavirus Vaccination on Diarrheal Hospitalizations in Children Aged
30
71. Bauch CT, Earn DJ. Vaccination and the theory of games. Proc Natl Acad Sci U S
A. 2004;101(36):13391-4. 72. Simon H. From substantive to procedural rationality, in HA Simon (ed.) Models of
bounded rationality: behavioural economics and business organization. . Cambridge, MA: MIT Press; 1976.
73. Fine PE, Clarkson JA. Individual versus public priorities in the determination of
optimal vaccination policies. Am J Epidemiol. 1986;124(6):1012-20. 74. WHO. Global Vaccine Safety. WHO vaccine reaction rates information sheets.
Available at: http://www.who.int/vaccine_safety/initiative/tools/vaccinfosheets/en/ [accessed 13th August 2018]. 2018.
75. Resnik D. The Ethics of Research with Human Subjects: Protecting People,
Advancing Science, Promoting Trust. 1 ed: Springer International Publishing; 2018. XIII, 316 p.
76. WHO. Global Advisory Committee on Vaccine Safety (GACVS). Available at:
http://www.who.int/vaccine_safety/initiative/communication/network/_gacvs/en/ [Accessed 31st August 2018].
http://www.who.int/vaccine_safety/initiative/tools/vaccinfosheets/en/http://www.who.int/vaccine_safety/initiative/communication/network/_gacvs/en/
31
3.0 Chapter 3 - Aim and objectives of thesis
3.1 Aim
The aim of this thesis is to estimate the potential mortality benefits (averted RVGE
deaths aged
32
4.0 Chapter 4 - Estimation of rotavirus deaths in children aged
33
4.3 Ethical approval
Appendix 3 (Chapter 4, S1 Table, Information about the data used for new analyses)
describes the ethical approvals obtained by all collaborators/co-authors involved in
this work. I did not seek LSHTM ethical approval for my contribution to the paper
because my analysis was based on publicly available datasets and published papers in
the public domain. I had no access to files with patient identifiable data and did not
analyse or have access to any primary databases.
RESEARCH ARTICLE
Estimating global, regional and national
rotavirus deaths in children aged
watery diarrhoea deaths. 97% (95% CI 9598%) of the U5 diarrhoea hospitalisations that
tested positive for rotavirus were entirely attributable to rotavirus. For all clinical syndromes
combined the rotavirus attributable fraction was 34% (95% CI 3136%). This increased by a
factor of 1.08 (95% CI 1.021.14) when the GEMS results were reanalysed using a more
sensitive molecular test.
Conclusions
We developed consensus on seven proposals for improving the quality and transparency of
future rotavirus mortality estimates.
Introduction
Rotavirus is a leading cause of diarrhoeal mortality in children less than five years old (U5),
but there is considerable disagreement about how many rotavirus deaths occur each year.
Recent estimates from different sources range from ~120,000 to ~215,000 [13]. Accurate
rotavirus mortality estimates help governments and donors prioritise public health interven-
tions and provide a basis for assessing the impact of immunization on mortality rates. Conflict-
ing estimates from different sources create confusion and can delay the introduction of
important diarrhoea mortality prevention measures, such as rotavirus vaccines.
In recent years, three groups have produced estimates of rotavirus deaths:
1. CHERGthe Child Health Epidemiology Reference Group of the World Health Organiza-
tion (WHO) and UNICEF. CHERG is now referred to as MCEEthe Maternal and Child
Epidemiology Estimation group;
2. GBDthe Global Burden of Disease Study, a collaboration led by the Institute for Health
Metrics and Evaluation (IHME); and,
3. WHO/CDCthe WHO and Centers for Disease Control and Prevention (joint estimates).
A meeting coordinated by WHO (Geneva, March 2015) facilitated the initial discussions on
the differences between the currently available rotavirus mortality estimates. This work builds
on a previous assessment of differences between CHERG and GBD estimates of all-cause U5
diarrhoea deaths [4]. Several gaps in the evidence were identified at an early stage in the pro-
cess, and one important task was to conduct new analyses to help bridge these gaps. First, rota-
virus is not associated clinically with acute bloody (dysenteric) diarrhoea and rarely with
persistent diarrhoea (of 14 days duration or more). As a result, many of the rotavirus-positive
proportions reported in hospital surveillance networks, and in the literature, exclude these
cases, and simply report the rotavirus-positive proportion among hospitalised children with
acute watery diarrhoea. If this proportion is applied to all episodes of diarrhoea resulting in
hospitalisation, it will result in overestimates. Second, there is very limited evidence to inform
whether the distribution of clinical syndromes for U5 diarrhoea hospitalisations (% acute
watery, % acute bloody, % persistent) is similar to, and thus a reasonable proxy for, the distri-
bution of clinical syndromes for U5 diarrhoea deaths. Most approaches assume that rotavirus-
positivity among diarrhoea hospitalisations is a reasonable proxy for rotavirus-positivity
among diarrhoea deaths. However, the two proportions are rarely reported in the same study
population. Third, to date there has been no explicit quantification of the rotavirus attributable
Estimating rotavirus deaths in children aged
fraction among U5 diarrhoea hospitalisations, or the extent to which that varies depending on
the type of diagnostic test used.
The aim of this manuscript is to compare the existing rotavirus mortality estimates, explain
the reasons for differences, provide evidence to inform key areas of uncertainty, and propose
improvements for future estimates.
Methods
We used a range of methods and sources of data. First, we compared existing estimates of U5
rotavirus deaths at the global, regional and national level and identified key differences in the
approaches used. Second, we used data from a large number of hospitals to estimate the pro-
portion of U5 diarrhoea hospitalisations that were acute watery, acute bloody and persistent.
Third, we used data from verbal autopsy studies to estimate the proportion of U5 diarrhoea
deaths that were acute watery, acute bloody and persistent. Fourth, we calculated the propor-
tion of U5 diarrhoea hospitalisations and U5 diarrhoea deaths that were rotavirus-positive in
each of the African and Asian sites included in the Global Enteric Multicenter Study (GEMS).
Fifth, we used data from GEMS to estimate the proportion of rotavirus-positive U5 diarrhoea
hospitalisations that were entirely attributable to rotavirus, and quantified the increase in the
rotavirus attributable fraction when a more sensitive molecular test was used to determine
rotavirus-positivity.
All data used in this study were anonymized prior to access and analysis. Please see support-
ing information (S1 Table) for details about institutional ethical approvals, and how and where
the data were collected.
U5 rotavirus deaths: Comparison of estimates from GBD, CHERG and
WHO/CDC
An independent reviewer (AC) compared the methods and data files published by the Global
Burden of Disease 2013 Study (GBD 2013) [1, 5], CHERG [2, 6] and WHO/CDC [3, 7]. GBD
provided a data file with country specific estimates of U5 rotavirus deaths [8].
We compared global, regional and national estimates of U5 deaths, U5 diarrhoea deaths
and U5 rotavirus deaths for the year 2013 using a standard list of 186 countries (S1 File).
CHERG did not report country estimates of U5 rotavirus deaths, so we multiplied country
estimates of U5 diarrhoea deaths for the year 2013 by regional estimates of the proportion of
U5 diarrhoea deaths due to rotavirus, as reported by CHERG for the year 2010. We removed
two countries from the GBD list (Taiwan, Palestine) and seven from the WHO/CDC list
(Cook Islands, Monaco, Nauru, Niue, Palau, St Kitts and Nevis, San Marino, Tuvalu) because
they did not appear in both GBD and WHO/CDC datasets. GBD, CHERG and WHO/CDC
used different classifications for grouping countries. For the purpose of this comparison exer-
cise, all countries were grouped using the WHO classification system i.e. AFRO, AMRO,
EMRO, EURO, SEARO, WPRO [9].
Clinical syndromes of U5 diarrhoea hospitalisations: Acute watery, acute
bloody, persistent
To estimate the proportion of U5 diarrhoea hospitalisations that were acute watery, acute
bloody and persistent, we used data from 84 hospitals in 9 countries:
1. 50 hospitals (5 in Indonesia, 42 in Rwanda and 3 in Zambia) from the WHO-coordinated
Global Sentinel Site Rotavirus Surveillance NetworkGRSN [10];
Estimating rotavirus deaths in children aged
2. 7 hospitals from the Indian National Hospital Rotavirus Surveillance NetworkNRSN
(Delhi, Hyderabad, Kolenchery, Ludhiana, Tirupati, Trichy and Vellore); and;
3. 27 hospitals included in the Global Enteric Multicenter StudyGEMS (1 in Bangladesh, 4
in India, 6 in Gambia, 10 in Kenya, 6 in Mozambique). Methods for recruiting and enroll-
ing moderate-to-severe diarrhoea (MSD) cases in GEMS have been described in detail else-
where [11]. We included 5 of the 7 GEMS sites in this particular analysis. Mali and Pakistan
were excluded because they rarely hospitalised children [12].
To be included, sites had to be major paediatric hospitals or district hospitals with100
children aged
more than 9 children with MSD were identified in a fortnight, only the first 9 children were
enrolled and tested for rotavirus; the remainder were recorded on a log and assumed to have
the same rotavirus-positive proportion as enrolled cases that were identified in the same fort-
night, age stratum (0-11m, 12-23m, 24-59m) and diarrhoea syndrome (acute watery, acute
bloody, persistent).
We also calculated the proportion of U5 deaths that: a) tested positive for rotavirus within 7
days of death; and, b) had diarrhoea coded as the first or second cause of death on their verbal
autopsy (VA) report. Rotavirus-positive children with a missing VA report (~20%) were
assumed to have the same cause-of-death breakdown as rotavirus-positive children with a VA
report.
For completeness, we also calculated the rotavirus-positive proportion among healthy con-
trols as well as MSD cases that were not admitted to hospital.
Proportion of rotavirus-positive U5 diarrhoea hospitalisations attributable
to rotavirus in GEMS
GEMS tested for a wide range of enteric pathogens in the stools of MSD cases and healthy
community controls without diarrhoea matched to cases by age, gender, and residence; con-
trols were enrolled within 14 days of the index case. GEMS also included information about
whether MSD cases were admitted to hospital or not.
We used multiple conditional logistic regression to calculate the odds ratio of rotavirus EIA
positivity in hospitalized MSD cases vs matched healthy controls adjusted for the presence of
other pathogens. All syndromes of diarrhoea were included. We then calculated the attribut-
able fraction (AF) as described by Bruzzi et al [15]. These methods were the same as those usedto estimate attributable fractions in the main GEMS analysis [12, 16]. However, we restricted
the analysis to hospitalised cases, thought to be a better proxy for estimating rotavirus-
attributable mortality than all MSD cases. We excluded Mozambique from all AF analyses due
to concerns about the quality of the EIA testing, and did not estimate individual AFs for Mali
and Pakistan because hospitalisation for diarrhoea was very rare in these sites.
Using these attributable fractions, which represent the fraction of hospitalised MSD cases
with disease attributable to rotavirus, we calculated the attributable fraction among the
exposed (AFe). The AFe represents the fraction of rotavirus positive cases who have disease
caused by rotavirus. The rotavirus-positive proportions used to derive the AF and AFe were
based only on the children with MSD that were tested for rotavirus. These were age-specific
(0-11m, 12-23m, 24-59m) and did not involve extrapolation to non-enrolled MSD cases.
Finally, we used previously described methods [17] to calculate the rotavirus attributable
fraction based on quantitative Polymerase Chain Reaction (qPCR). We restricted the analysis
to a subset of 721 hospital cases and matched controls, and calculated the AF for all country
sites combined, excluding Mozambique. To quantify the test performance of EIA compared to
qPCR, we repeated this analysis for EIA test results, and calculated the ratio between the two
attributable fractions. All syndromes of diarrhoea were included. Confidence intervals were
calculated by bootstrapping with 1000 iterations.
Results
Comparison exercise
GBD produce their own estimates of U5 deaths [18], whereas CHERG and WHO/CDC use
U5 deaths from the UN Inter-agency Group for Child Mortality Estimation (IGME)[19]. Both
GBD and IGME estimate approximately 6.3 million U5 deaths globally in 2013 (Table 1) but
Estimating rotavirus deaths in children aged
some important differences exist at country/regional levels e.g. ~739,000 (GBD) vs ~845,000
(IGME) in the Eastern Mediterranean Region (EMRO). The main methodological differences
between GBD and IGME have been described in detail elsewhere and include the choice of
data points selected (vital registration, census and household surveys) and fitting methods
used [20].
GBD and CHERG produce their own estimates of the proportion of U5 deaths due to diar-
rhoea [5, 6]; WHO/CDC use the CHERG estimates. GBD and CHERG estimated that 89% of
U5 deaths were caused by diarrhoea at the global level in the year 2013 (Table 1). Differences
in GBD and CHERG estimates for the South East Asia (SEARO) region (6% vs 10%) are driven
by differences in estimates for India (6% vs 10%) where U5 diarrhoea deaths are ~80,000 vs
~140,000 respectively (Table 1). In other regions there is more agreement. Estimates for the
African (AFRO) region are consistent overall (10% vs 10%) but there are still large differences
at country level e.g. Zimbabwe (Fig 1).
Methodological differences between GBD and CHERG have been described in detail else-
where [4]. In brief, CHERG excluded verbal autopsy studies that only investigated a single
cause of death and data points from incomplete vital registration systems in higher mortality
Table 1. Comparison of CHERG, GBD and WHO/CDC estimates of U5 deaths, U5 diarrhoea deaths and U5 rotavirus deaths in the year 2013 by
WHO region, and for selected large countries.
GLOBAL AFRO AMRO EMRO EURO SEARO WPRO Bangladesh DR
Congo
India Indonesia
U5 deaths
UN (IGME) used by
CHERG
6,282,254 2,977,576 227,475 845,286 136,850 1,700,178 394,889 129,433 319,977 1,340,055 136,371
GBD 6,271,643 3,164,861 248,643 738,702 130,573 1,604,028 384,836 128,228 340,416 1,249,673 148,807
WHO/CDC - - - - - - - - - - -
Proportion of U5 deaths
due to diarrhoea
CHERG 0.09 0.10 0.04 0.10 0.04 0.10 0.06 0.06 0.11 0.10 0.06
GBD 0.08 0.10 0.05 0.12 0.03 0.06 0.02 0.01 0.17 0.06 0.06
WHO/CDC - - - - - - - - - - -
U5 diarrhoea deaths
CHERG 577,508 293,289 9,297 84,592 5,689 162,298 22,344 8,298 33,730 140,451 7,505
GBD 519,485 312,297 11,923 88,071 3,694 94,574 8,926 1,715 57,344 80,188 8,694
WHO/CDC 577,508 293,289 9,297 84,592 5,689 162,298 22,344 8,298 33,730 140,451 7,505
Proportion of U5
diarrhoea deaths due to
Rotavirus
CHERG 0.27 0.27 0.23 0.31 0.26 0.26 0.33 0.26 0.27 0.26 0.26
GBD 0.24 0.24 0.18 0.18 0.26 0.27 0.42 0.12 0.13 0.26 0.37
WHO/CDC 0.37 0.39 0.26 0.36 0.31 0.35 0.43 0.33 0.40 0.34 0.50
U5 rotavirus diarrhoea
deaths
CHERG 157,398 78,601 2,176 26,477 1,473 41,386 7,284 2,116 9,040 35,815 1,914
GBD 122,322 73,758 2,178 15,984 976 25,637 3,790 202 7,523 21,205 3,176
WHO/CDC 215,757 115,023 2,455 30,577 1,752 56,287 9,664 2,723 13,526 47,082 3,771
Region and global estimates may differ from official WHO/CDC, CHERG and GBD estimates because a standard set of countries and regions was used and
no rounding was done prior to aggregation.
https://doi.org/10.1371/journal.pone.0183392.t001
Estimating rotavirus deaths in children aged
settings. GBD included these data points and adjusted for missing data. GBD also included
unpublished data points obtained under third party data use agreements whereas CHERG
only use publicly available data points [21].
All three groups produce their own estimates of the proportion of U5 diarrhoea deaths that
are attributable to rotavirus. For the year 2013, the global proportions were 24% (GBD), 27%
(CHERG) and 37% (WHO/CDC). These correspond to 122,322 (GBD), 157,398 (CHERG)
and 215,757 (WHO/CDC) U5 rotavirus deaths (Table 1). Fig 2 shows the extent of variation in
the fraction of diarrhoea deaths attributed to rotavirus across countries within each WHO
region. There are large differences in some countries; for example, in DR Congo the propor-
tions are 13% (GBD), 27% (CHERG) and 40% (WHO/CDC).
The three groups used different methods to:
1. select data points (rotavirus-positive proportions);
2. extrapolate data points to individual countries;
3. account for rotavirus vaccine coverage;
4. convert rotavirus-positive proportions to rotavirus attributable fractions; and,
5. calculate uncertainty ranges.
A more detailed description of these differences can be found in the supporting informa-
tion (S1 Appendix).
Fig 1. Country-level differences in GBD vs CHERG estimates of the proportion of U5 deaths due to
diarrhoea in the year 2013 by WHO region.
https://doi.org/10.1371/journal.pone.0183392.g001
Estimating rotavirus deaths in children aged
Clinical syndromes for U5 diarrhoea hospitalisation
Table 2 shows the distribution of clinical syndromes for U5 diarrhoea hospitalisations for vari-
ous sites in Africa and Asia. A meta-analysis including data from all GRSN, NRSN and GEMS
sites suggests that acute watery diarrhoea was associated with 87% (95% CI 8390%) of U5
diarrhoea hospitalisations (Fig 3) but there was substantial evidence for heterogeneity (I-
squared 99.08%, p = 0.00) between the studies. The GEMS site in Bangladesh (Mirzapur) had a
very high rate of acute bloody diarrhoea for reasons that are not clear.
Clinical syndromes for U5 diarrhoea deaths
Table 2 shows the distribution of clinical syndromes for U5 diarrhoea deaths. A meta-analysis
suggests that acute watery diarrhoea was associated with 65% (95% CI 5774%) of U5 diar-
rhoea deaths (Fig 4) but again there was substantial evidence for heterogeneity between the
studies (I-squared 92.06%, p = 0.00). In four of the nine countries with verbal autopsy data, the
clinical syndromes of diarrhoea deaths were compared for those who died in any type of health
facility and those who died in the home, as reported by the family respondent. Most of deaths
were in the home (Cameroon 70%, Malawi 50%, Niger 86% and Nigeria 78%) but the distribu-
tion of acute watery, acute bloody and persistent diarrhoea was similar irrespective of the place
of death (Kalter, personal communication).
Rotavirus-positive proportion in U5 diarrhoea hospitalisations and U5
diarrhoea deaths in GEMS
For all GEMS sites combined (excluding Mozambique), rotavirus was detected (EIA-positive)
in 44% of acute watery U5 diarrhoea hospitalisations (55% in Asia, and 32% in Africa)
Fig 2. Country-level variation in the fraction of U5 diarrhoea deaths due to rotavirus in the year 2013
by source of estimates and by WHO region.
https://doi.org/10.1371/journal.pone.0183392.g002
Estimating rotavirus deaths in children aged
Table 2. Number and proportion of acute watery, acute bloody and persistent cases among U5 diarrhoea hospitalisations and U5 diarrhoea deaths
in various settings before rotavirus vaccine introduction.
Source Study
location
Study type Study
period
Diarrhoea
outcome
Age Total
n
Acute
Watery
n
Acute
Bloody
n
Persis-tent*n
Acute
Watery
%
Acute
Bloody
%
Persis-tent
%
Clinical syndromes of U5 diarrhoea hospitalisations
GRSN Indonesia Surveillance
hospitals (n = 5)
201415 Inpatients
Fig 3. Meta-analysis showing the proportion of U5 diarrhoea hospitalisations associated with acute
watery diarrhoea (AWD) for selected sites in Africa and Asia.
https://doi.org/10.1371/journal.pone.0183392.g003
Fig 4. Meta-analysis showing the proportion of U5 diarrhoea deaths associated with acute watery
diarrhoea (AWD) for selected sites in Africa and Asia.
https://doi.org/10.1371/journal.pone.0183392.g004
Estimating rotavirus deaths in children aged
(Table 3). When all clinical syndromes of diarrhoea were included, the rotavirus-positive pro-
portion was 38% (44% in Asia; 30% in Africa).
Rotavirus was detected (EIA-positive) in 32% (12/37) of children aged
Proportion of rotavirus-positive U5 diarrhoea hospitalisations attributable
to rotavirus in GEMS
The AFe value (equivalent to the rotavirus attributable fraction among rotavirus-positive U5
diarrhoea hospitalisations) was 0.97 (95% CI 0.950.98) for all included GEMS sites (Table 4)
and all diarrhoea syndromes combined.
Using qPCR instead of EIA for rotavirus detection increased the AF by a factor of 1.08
(95% CI 1.021.14).
Proposed improvements
We propose a number of improvements for consideration by all groups involved in the devel-
opment of future rotavirus mortality estimates.
Reporting a standard set of minimum variables to describe all input data
points
Previous comparison exercises have stressed the need for input data points to be made avail-
able at the time estimates are published [4, 20]. Recent Guidelines for Accurate and Transpar-
ent Health Estimates Reporting (GATHER) have recommended publication of a spreadsheet
table with details about the data points used to inform estimates [22]. These guidelines do not
provide explicit guidance on the variables that should be reported. We suggest that the follow-
ing standard set of minimum variables should be reported: (a) author/reference; (b) country;
(c) sub-national location; (d) data collection period; (e) age range; (f) type of study; (g) type of
diagnostic test; (h) number of enteric pathogens tested; (i) inpatient/outpatient; (j) pre/post
Table 4. Rotavirus positive proportion, attributable fraction (AF) and attributable fraction in the exposed (AFe) for MSD cases
implementation of rotavirus vaccine in the public sector, or preferably a more precise estimate
of rotavirus vaccine coverage with details about the source of sub-national or national coverage
data used; (k) type of clinical syndrome e.g. acute watery, all syndromes; (l) included/excluded
in final estimates; (m) justification if excluded; (n) rotavirus-positive proportion (unadjusted);
(o) rotavirus-positive proportion (adjusted); and, (p) description of adjustment applied. Inclu-
sion and exclusion criteria should be clearly documented, and any exclusions applied after
data extraction should be justified using a clearly defined framework for evaluating data qual-
ity and outliers.
Annual online publication of WHO surveillance data points in
spreadsheet format
A spreadsheet table should be published annually on the WHO web site to allow for potential
inclusion of GRSN data by all groups in future estimates. At a minimum, rotavirus-positive
proportions
derived exclusively from acute watery U5 diarrhoea hospitalisations should be adjusted to
account for the proportion of total U5 diarrhoea hospitalisations that are acute watery. If the
rotavirus-positive proportion (r) is not reported for all clinical syndromes combined, then theequation r = ab + c(1 b) can be used, where a is the rotavirus-positive proportion amongacute watery U5 diarrhoea hospitalisations, b is the proportion of total U5 diarrhoea hospitali-sations that are acute watery, and c is the rotavirus-positive proportion among acute bloodyand persistent U5 diarrhoea hospitalisations combined. In the absence of local data to inform
parameter b, our analysis shows that acute watery diarrhoea is likely to be responsible for nomore than 87% (95% CI 8390%) of U5 diarrhoea hospitalisations. The true value of b is likelyto be lower because all data points included in the meta-analysis under-estimated the role of
persistent diarrhoea. Given that rotavirus is not associated clinically with acute bloody or per-
sistent diarrhoea, the value of parameter c is likely to be at least ~3% based on the rotavirus-positivity observed in healthy controls in GEMS.
Accounting for uncertainty in the steps used to convert rotavirus-positive
proportions into rotavirus-attributable fractions
The frequent asymptomatic carriage of many pathogens in the stools of healthy controls neces-
sitates the calculation of attributable fractions. To estimate the proportion of rotavirus-positive
cases that are attributable only to rotavirus, the population attributable fraction estimated by
the equation r AFe can be used, where r is the rotavirus-positive proportion reported amongU5 diarrhoea hospitalisations (all syndromes combined), and AFe is the rotavirus-attributablefraction among rotavirus-positive U5 diarrhoea hospitalisations (all syndromes combined).
Because it is rare for diarrhoea surveillance studies to include diarrhoea-free controls, very few
studies allow calculation of AFe. GEMS does include diarrhoea-free controls so permits thiscalculation; our new analysis of GEMS calculated the AFe to be 0.97 (95% CI 0.950.98). This
value was relatively consistent across all GEMS sites where it could be reported (Bangladesh,
India, Gambia, Kenya). This suggests that rotavirus is the attributable cause in almost all U5
rotavirus-positive diarrhoea hospitalisations. In a separate, related analysis, the rotavirus
attributable fraction was shown to increase by a factor of 1.08 (95% 1.021.14) when the more
sensitive qPCR test was used. This is similar (albeit slightly larger) than the adjustment made
to r to account for AFe, so both adjustments could reasonably be excluded, and this wouldhave a limited impact on central estimates of U5 rotavirus deaths. However, adjustments
applied to some pathogens and not others, would lead to inconsistent reporting of central esti-
mates (and uncertainty intervals) across enteric pathogens. These adjustments, and their
uncertainty, should therefore be reflected in future estimates for all enteric pathogens, includ-
ing rotavirus.
Further research into the clinical syndromes of U5 diarrhoea deaths, and
the real-world impact of rotavirus vaccines on those deaths
To date, all groups have assumed that the proportion of U5 diarrhoea hospitalisations caused
by rotavirus is a reasonable proxy for the proportion of U5 diarrhoea deaths caused by rotavi-
rus. This approach has been taken because hospitalisation is thought to be a good proxy for
diarrhoea that is sufficiently severe to lead to death. Two aspects of our analysis suggest this
assumption may lead to over-estimates of the number of U5 rotavirus deaths. First, we esti-
mate that acute watery diarrhoea is associated with 87% of diarrhoea hospitalisations but only
65% of U5 diarrhoea deaths. Higher case fatality ratios (CFR) have been reported for acute
bloody and persistent diarrhoea than acute watery diarrhoea [23] but more evidence on the
fatality of different syndromes is needed to corroborate this. In addition, the analysis of
Estimating rotavirus deaths in children aged
diarrhoeal deaths relied on verbal autopsy reports which may be prone to recall bias, and our
analysis of diarrhoea hospitalisations only included those that became persistent after admis-
sion. Second, rotavirus was detected in a higher proportion of U5 acute watery diarrhoea hos-
pitalisations than U5 acute watery diarrhoea deaths in GEMS (44% vs 28%). Thus, among
children that had access to treatment, rotavirus was estimated to be less fatal than other causes
of acute watery diarrhoea. However, more evidence is needed on the effect of treatment on the
proportion of acute watery diarrhoea deaths due to rotavirus; in communities without access
to treatment services, rotavirus may represent a larger proportion of acute watery diarrhoea
deaths. Another explanation for the lower rotavirus-positivity among U5 acute watery diar-
rhoea deaths, is that the number of deaths captured in the 7 days after enrolment (n = 37) were
too few to make a reliable assessment. Longer follow-up periods allow more deaths to be
included but it then becomes increasingly difficult to ascertain whether children who were
rotavirus-positive at the time of enrolment were still rotavirus-positive at the time of death,
and whether cases that were negative at enrolment had a new rotavirus episode prior to death.
Further evidence is needed from other geographical locations on the distribution of clinical
syndromes among U5 diarrhoea hospitalisations and deaths. This should include a more accu-
rate assessment of the role of persistent diarrhoea among U5 diarrhoea hospitalisations. More
importantly, efforts should be made to accurately capture the real-world impact of rotavirus
vaccines on U5 diarrhoea deaths in early introducing countries. This will provide critical
insights into the true contribution of rotavirus to U5 diarrhoea deaths in different locations.
Presenting and incorporating the uncertainty in parameters used to
derive U5 rotavirus deaths
The uncertainty interval around the central estimates of U5 rotavirus deaths should be explic-
itly defined (e.g. the type of confidence or prediction interval) and should incorporate uncer-
tainty in each of the three core parameters (number of U5 deaths, % due to diarrhoea, % due
Fig 5. Global estimates of the number of rotavirus deaths
to rotavirus) as well as any other parameters used to adjust the original input data points e.g.
the parameters used to convert rotavirus-positive proportions into rotavirus-attributable
fractions.
Conclusion
There is considerable disagreement between global estimates of U5 rotavirus deaths, but it is
encouraging to note that estimates are converging over time, at least in absolute terms (Fig 5).
The aim of this analysis was not to recommend a single set of best estimates, but rather to
explain the reasons for differences, provide evidence to inform key areas of uncertainty, and
propose improvements for future estimates. Updates to GBD [24] and CHERG (now MCEE)
estimates were already well advanced during the course of this comparison study, and further
convergence is expected. The suggested improvements presented in this manuscript should be
incorporated, as far as possible, into future rotavirus mortality estimates. This is likely to be an
iterative and evolving process as new evidence emerges over time.
Supporting information
S1 Table. Information about the data used for new analyses.
(DOCX)
S1 Appendix. Further details on the comparison of rotavirus mortality estimates from
GBD, CHERG and WHO/CDC.
(DOCX)
S1 File. Country-level dataset used to compare CHERG, GBD and WHO/CDC estimates of
U5 deaths, U5 diarrhoea deaths and U5 rotavirus deaths in the year 2013.
(XLSM)
Acknowledgments
Disclaimer: The findings and conclusions of this report are those of the authors and do not
necessarily represent the official position of the Centers for Disease Control and Prevention
(CDC).
GRSN author group: Yati Soenarto (University of Gadjah Mada, Yogyakarta, Indonesia);
Celse Rugambwa (World Health Organization, Kigali, Rwanda); Evans Mpabalwani (Depart-
ment of Paediatrics & Child Health, Lusaka, Zambia); Jason Mwenda (World Health Organi-
zation, Brazzaville, Republic of Congo), Jill Murray, Adam Cohen (World Health
Organization, Geneva, Switzerland).
We acknowledge the many country-level collaborators involved in the GRSN, NRSN and
GEMS study sites. We acknowledge Ximena Riveros and Ana Maria Henao-Restrepo from
WHO Initiative for Vaccine Research, who helped to organise the initial meeting and bring
together the various rotavirus disease experts. We thank Ulla Griffiths and Mark Jit for provid-
ing useful comments on the paper.
Author Contributions
Formal analysis: Andrew Clark, Robert Black, Jacqueline Tate, Anna Roose, Karen Kotloff,
Diana Lam, William Blackwelder, Gagandeep Kang, James Platts-Mills, Colin Sanderson.
Investigation: Andrew Clark, Robert Black, Jacqueline Tate, Anna Roose, Karen Kotloff,
Diana Lam, William Blackwelder, Gagandeep Kang, James Platts-Mills, Colin Sanderson.
Estimating rotavirus deaths in children aged
Methodology: Andrew Clark, Robert Black, Jacqueline Tate, Anna Roose, Karen Kotloff,
Diana Lam, William Blackwelder, Umesh Parashar, Claudio Lanata, Gagandeep Kang,
Christopher Troeger, James Platts-Mills, Ali Mokdad, Colin Sanderson, Laura Lamberti,
Myron Levine, Mathuram Santosham, Duncan Steele.
Writing original draft: Andrew Clark.
Writing review & editing: Andrew Clark, Robert Black, Jacqueline Tate, Anna Roose, Karen
Kotloff, Diana Lam, William Blackwelder, Umesh Parashar, Claudio Lanata, Gagandeep
Kang, Christopher Troeger, James Platts-Mills, Ali Mokdad, Colin Sanderson, Laura Lam-
berti, Myron Levine, Mathuram Santosham, Duncan Steele.
References1. Global, regional, and national age-sex specific all-cause and cause-specific mortality for 240 causes of
death, 19902013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet. 2015;
385(9963):11771. https://doi.org/10.1016/S0140-6736(14)61682-2 PMID: 25530442
2. Lanata CF, Fischer-Walker CL, Olascoaga AC, Torres CX, Aryee MJ, Black RE. Global causes of diar-
rheal disease mortality in children
17. Liu J, Platts-Mills JA, Juma J, Kabir F, Nkeze J, Okoi C, et al. Use of quantitative molecular diagnostic
methods to identify causes of diarrhoea in children: a reanalysis of the GEMS case-control study. Lan-
cet. 2016; 388(10051):1291301. https://doi.org/10.1016/S0140-6736(16)31529-X PMID: 27673470
18. IHME. Mortality estimation process. https://vizhub.healthdata.org/mortality/ [accessed 19th May 2016].
19. IGME. UN Inter-agency Group for Child Mortality Estimation. http://www.childmortality.org/ [accessed
18th March 2016].
20. Alkema L, You D. Child mortality estimation: a comparison of UN IGME and IHME estimates of levels
and trends in under-five mortality rates and deaths. PLoS Med. 2012; 9(8):e1001288. https://doi.org/10.
1371/journal.pmed.1001288 PMID: 22952434
21. Vos Theo B R, Phillips David E, Lopez Alan D, Murray Christopher J L. Authors Reply to Correspon-
denceCauses of child death: comparison of MCEE and GBD 2013 estimates. Liu Li, Black Robert E,
Cousens Simon, Mathers Colin, Lawn Joy E, Hogan Daniel R. Published Online May 23, 2015 http://dx.
doi.org/10.1016/S0140-6736(15)60663-8.
22. Stevens GA, Alkema L, Black RE, Boerma JT, Collins GS, Ezzati M, et al. Guidelines for Accurate and
Transparent Health Estimates Reporting: the GATHER statement. Lancet. 2016.
23. Bhandari N, Bhan MK, Sazawal S. Mortality associated with acute watery diarrhea, dysentery and per-
sistent diarrhea in rural North India. Acta Paediatr. 1992; 81 Suppl 381:36.
24. Mortality GBD Causes of Death C. Global, regional, and national life expectancy, all-cause mortality,
and cause-specific mortality for 249 causes of death, 19802015: a systematic analysis for the Global
Burden of Disease Study 2015. Lancet. 2016; 388(10053):1459544. https://doi.org/10.1016/S0140-
6736(16)31012-1 PMID: 27733281
Estimating rotavirus deaths in children aged
52
5.0 Chapter 5 - Estimation of rotavirus age distributions in children aged
53
Global review of the age distribution of rotavirus disease in children aged
54
Abstract
Background: The impact of live oral rotavirus vaccines could be improved by adjusting the
schedules, but in published age distributions of rotavirus gastroenteritis (RVGE) the age bands
are too broad to allow a detailed investigation of the potential gains.
Methods: We sought datasets that could provide age distributions of rotavirus-positive
community cases, clinic visits, hospital admissions, emergency visits and deaths among
children
55
Introduction
Rotavirus gastroenteritis (RVGE) is estimated to cause around 200,000 child deaths each year
(1). Over half of the countries in the world now include live oral rotavirus vaccines in their
national immunization programmes (2). There are three vaccines licensed for global use
(Rotarix - GSK, RotaTeq - Merck & Co., and ROTAVAC - Bharat Biologicals), others
for national use (e.g. in Vietnam, China and India) and several others in the pipeline, including
neonatal and non-replicating injectable vaccines (3). Randomised controlled trials (RCTs)
have reported high vaccine efficacy (~90%) against severe RVGE in low mortality countries
but modest efficacy (~50%) in higher mortality settings (4). Alternative schedules are being
considered to increase their impact. A neonatal vaccine has had promising results in
Indonesia(5), and some studies have evaluated the potential of a booster dose given at around
9-12 months of age (6, 7). Several studies and surveillance systems have collected information
on RVGE age distributions but much of it is unpublished or has been published in age bands
that are too broad to allow a detailed assessment of the potential impact of alternative rotavirus
vaccination schedules. More granular age distributions would also help to quantify the number
of RVGE cases expected to occur at specific ages, so that changes can be monitored after
vaccination. More generally, there is a need to update the global evidence on RVGE age
distributions, compare them between countries and regions, and establish a reliable method
for extrapolating them to countries without data. An unpublished review was conducted in
2012 (8) but this did not include the large multi-country Global Rotavirus Surveillance
Network (GRSN) database (9), and several pivotal multi-country studies have also been
published since (10-12).
In this paper we aim to estimate granular age distributions of rotavirus disease outcomes in
children aged
56
Methods
Ethical Approval
This study was approved by the ethical committee of the London School of Hygiene and
Tropical Medicine (LSHTM); ethics reference 14398. All authors and countries gave their
consent to analyse and publish the data.
Search strategy and study selection
We sought country datasets containing counts of rotavirus-positive disease in children aged
57
distributions; and, papers without an accessible full text link. Two independent reviewers
(MHA, CL) screened abstracts and any ambiguity was resolved by a third reviewer (AC). A
letter was sent by email to the investigators of all studies identified in the systematic review.
Investigators were asked to provide anonymised data or complete a standard data extraction
table with counts by week of age up to 5.0 years. If the investigators did not respond before
the end of August 2017 and no other study was available for that country, we extracted the
age distribution reported in the publication. We included all country datasets that were
obtained from a previously unpublished literature and database search conducted by
Sanderson et al in 2012(8). This included articles published between 1990 and 2011.
All country datasets were combined into a central database with a standard format and list of
variables and analysed together with the GRSN datasets. We cross-checked datasets identified
through the literature search and GRSN to avoid data duplication. Prior to analysing the
datasets, we excluded studies that included fewer than 35 RVGE events, had known concerns
about EIA quality, had fewer than three age bands
58
reported by the UNPOP 2017 Revision(14). We grouped all datasets according to the under 5
mortality quintile of the country concerned, and calculated the median age and median best-
fitting parameters for each stratum. We also ran a series of regression analyses to explore
which combinations of variables would best predict the median age and parameters of the
chosen parametric distribution. To compare differences in rotavirus disease presentations we
plotted the full set of median ages reported for a given presentation against their respective
2010-2015 under-five mortality rates. We fit a least-squares line of best fit for each
presentation, reported the R-squared value and compared the best-fitting lines.
We used ArcGIS mapping software to display the median age of rotavirus hospitalisation
estimated for each country in the world. If more than a single dataset was available for a
country, we calculated the median age and median best-fitting parameters of all datasets for
that country. If no dataset was available, we assigned the median age of the countrys
corresponding mortality stratum.
Results
We identified 117 pre-vaccination datasets with rotavirus-positive events among children
59
example, in the very high child mortality stratum, the median age ranged from 29 weeks (IQR:
19-46) in Zambia to 50 weeks in Ethiopia (IQR: 30-81). Similarly, in the low/very low
mortality stratum, the median age ranged from 35 weeks (IQR: 19-64) in France to 101 weeks
(IQR: 65-157) in Ukraine.
Globally, most countries with a low median age were in Africa (Figure 3). In general, the
median age of rotavir